mailinfo.c

   1 #include "cache.h"
   2 #include "config.h"
   3 #include "utf8.h"
   4 #include "strbuf.h"
   5 #include "mailinfo.h"
   6
   7 static void cleanup_space(struct strbuf *sb)
   8 {
   9         size_t pos, cnt;
  10         for (pos = 0; pos < sb->len; pos++) {
  11                 if (isspace(sb->buf[pos])) {
  12                         sb->buf[pos] = ' ';
  13                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  14                         strbuf_remove(sb, pos + 1, cnt);
  15                 }
  16         }
  17 }
  18
  19 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  20 {
  21         struct strbuf *src = name;
  22         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  23                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  24                 src = email;
  25         else if (name == out)
  26                 return;
  27         strbuf_reset(out);
  28         strbuf_addbuf(out, src);
  29 }
  30
  31 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  32 {
  33         /* John Doe <johndoe> */
  34
  35         char *bra, *ket;
  36         /* This is fallback, so do not bother if we already have an
  37          * e-mail address.
  38          */
  39         if (mi->email.len)
  40                 return;
  41
  42         bra = strchr(line->buf, '<');
  43         if (!bra)
  44                 return;
  45         ket = strchr(bra, '>');
  46         if (!ket)
  47                 return;
  48
  49         strbuf_reset(&mi->email);
  50         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  51
  52         strbuf_reset(&mi->name);
  53         strbuf_add(&mi->name, line->buf, bra - line->buf);
  54         strbuf_trim(&mi->name);
  55         get_sane_name(&mi->name, &mi->name, &mi->email);
  56 }
  57
  58 static const char *unquote_comment(struct strbuf *outbuf, const char *in)
  59 {
  60         int c;
  61         int take_next_literally = 0;
  62
  63         strbuf_addch(outbuf, '(');
  64
  65         while ((c = *in++) != 0) {
  66                 if (take_next_literally == 1) {
  67                         take_next_literally = 0;
  68                 } else {
  69                         switch (c) {
  70                         case '\\':
  71                                 take_next_literally = 1;
  72                                 continue;
  73                         case '(':
  74                                 in = unquote_comment(outbuf, in);
  75                                 continue;
  76                         case ')':
  77                                 strbuf_addch(outbuf, ')');
  78                                 return in;
  79                         }
  80                 }
  81
  82                 strbuf_addch(outbuf, c);
  83         }
  84
  85         return in;
  86 }
  87
  88 static const char *unquote_quoted_string(struct strbuf *outbuf, const char *in)
  89 {
  90         int c;
  91         int take_next_literally = 0;
  92
  93         while ((c = *in++) != 0) {
  94                 if (take_next_literally == 1) {
  95                         take_next_literally = 0;
  96                 } else {
  97                         switch (c) {
  98                         case '\\':
  99                                 take_next_literally = 1;
 100                                 continue;
 101                         case '"':
 102                                 return in;
 103                         }
 104                 }
 105
 106                 strbuf_addch(outbuf, c);
 107         }
 108
 109         return in;
 110 }
 111
 112 static void unquote_quoted_pair(struct strbuf *line)
 113 {
 114         struct strbuf outbuf;
 115         const char *in = line->buf;
 116         int c;
 117
 118         strbuf_init(&outbuf, line->len);
 119
 120         while ((c = *in++) != 0) {
 121                 switch (c) {
 122                 case '"':
 123                         in = unquote_quoted_string(&outbuf, in);
 124                         continue;
 125                 case '(':
 126                         in = unquote_comment(&outbuf, in);
 127                         continue;
 128                 }
 129
 130                 strbuf_addch(&outbuf, c);
 131         }
 132
 133         strbuf_swap(&outbuf, line);
 134         strbuf_release(&outbuf);
 135
 136 }
 137
 138 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
 139 {
 140         char *at;
 141         size_t el;
 142         struct strbuf f;
 143
 144         strbuf_init(&f, from->len);
 145         strbuf_addbuf(&f, from);
 146
 147         unquote_quoted_pair(&f);
 148
 149         at = strchr(f.buf, '@');
 150         if (!at) {
 151                 parse_bogus_from(mi, from);
 152                 goto out;
 153         }
 154
 155         /*
 156          * If we already have one email, don't take any confusing lines
 157          */
 158         if (mi->email.len && strchr(at + 1, '@'))
 159                 goto out;
 160
 161         /* Pick up the string around '@', possibly delimited with <>
 162          * pair; that is the email part.
 163          */
 164         while (at > f.buf) {
 165                 char c = at[-1];
 166                 if (isspace(c))
 167                         break;
 168                 if (c == '<') {
 169                         at[-1] = ' ';
 170                         break;
 171                 }
 172                 at--;
 173         }
 174         el = strcspn(at, " \n\t\r\v\f>");
 175         strbuf_reset(&mi->email);
 176         strbuf_add(&mi->email, at, el);
 177         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 178
 179         /* The remainder is name.  It could be
 180          *
 181          * - "John Doe <john.doe@xz>"                   (a), or
 182          * - "john.doe@xz (John Doe)"                   (b), or
 183          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 184          *
 185          * but we have removed the email part, so
 186          *
 187          * - remove extra spaces which could stay after email (case 'c'), and
 188          * - trim from both ends, possibly removing the () pair at the end
 189          *   (cases 'a' and 'b').
 190          */
 191         cleanup_space(&f);
 192         strbuf_trim(&f);
 193         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 194                 strbuf_remove(&f, 0, 1);
 195                 strbuf_setlen(&f, f.len - 1);
 196         }
 197
 198         get_sane_name(&mi->name, &f, &mi->email);
 199 out:
 200         strbuf_release(&f);
 201 }
 202
 203 static void handle_header(struct strbuf **out, const struct strbuf *line)
 204 {
 205         if (!*out) {
 206                 *out = xmalloc(sizeof(struct strbuf));
 207                 strbuf_init(*out, line->len);
 208         } else
 209                 strbuf_reset(*out);
 210
 211         strbuf_addbuf(*out, line);
 212 }
 213
 214 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 215  * to have enough heuristics to grok MIME encoded patches often found
 216  * on our mailing lists.  For example, we do not even treat header lines
 217  * case insensitively.
 218  */
 219
 220 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 221 {
 222         const char *ends, *ap = strcasestr(line, name);
 223         size_t sz;
 224
 225         strbuf_setlen(attr, 0);
 226         if (!ap)
 227                 return 0;
 228         ap += strlen(name);
 229         if (*ap == '"') {
 230                 ap++;
 231                 ends = "\"";
 232         }
 233         else
 234                 ends = "; \t";
 235         sz = strcspn(ap, ends);
 236         strbuf_add(attr, ap, sz);
 237         return 1;
 238 }
 239
 240 static int has_attr_value(const char *line, const char *name, const char *value)
 241 {
 242         struct strbuf sb = STRBUF_INIT;
 243         int rc = slurp_attr(line, name, &sb) && !strcasecmp(sb.buf, value);
 244         strbuf_release(&sb);
 245         return rc;
 246 }
 247
 248 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 249 {
 250         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 251         strbuf_init(boundary, line->len);
 252
 253         mi->format_flowed = has_attr_value(line->buf, "format=", "flowed");
 254         mi->delsp = has_attr_value(line->buf, "delsp=", "yes");
 255
 256         if (slurp_attr(line->buf, "boundary=", boundary)) {
 257                 strbuf_insert(boundary, 0, "--", 2);
 258                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 259                         error("Too many boundaries to handle");
 260                         mi->input_error = -1;
 261                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 262                         return;
 263                 }
 264                 *(mi->content_top) = boundary;
 265                 boundary = NULL;
 266         }
 267         slurp_attr(line->buf, "charset=", &mi->charset);
 268
 269         if (boundary) {
 270                 strbuf_release(boundary);
 271                 free(boundary);
 272         }
 273 }
 274
 275 static void handle_content_transfer_encoding(struct mailinfo *mi,
 276                                              const struct strbuf *line)
 277 {
 278         if (strcasestr(line->buf, "base64"))
 279                 mi->transfer_encoding = TE_BASE64;
 280         else if (strcasestr(line->buf, "quoted-printable"))
 281                 mi->transfer_encoding = TE_QP;
 282         else
 283                 mi->transfer_encoding = TE_DONTCARE;
 284 }
 285
 286 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 287 {
 288         struct strbuf *content_top = *(mi->content_top);
 289
 290         return ((content_top->len <= line->len) &&
 291                 !memcmp(line->buf, content_top->buf, content_top->len));
 292 }
 293
 294 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 295 {
 296         size_t at = 0;
 297
 298         while (at < subject->len) {
 299                 char *pos;
 300                 size_t remove;
 301
 302                 switch (subject->buf[at]) {
 303                 case 'r': case 'R':
 304                         if (subject->len <= at + 3)
 305                                 break;
 306                         if ((subject->buf[at + 1] == 'e' ||
 307                              subject->buf[at + 1] == 'E') &&
 308                             subject->buf[at + 2] == ':') {
 309                                 strbuf_remove(subject, at, 3);
 310                                 continue;
 311                         }
 312                         at++;
 313                         break;
 314                 case ' ': case '\t': case ':':
 315                         strbuf_remove(subject, at, 1);
 316                         continue;
 317                 case '[':
 318                         pos = strchr(subject->buf + at, ']');
 319                         if (!pos)
 320                                 break;
 321                         remove = pos - subject->buf + at + 1;
 322                         if (!mi->keep_non_patch_brackets_in_subject ||
 323                             (7 <= remove &&
 324                              memmem(subject->buf + at, remove, "PATCH", 5)))
 325                                 strbuf_remove(subject, at, remove);
 326                         else {
 327                                 at += remove;
 328                                 /*
 329                                  * If the input had a space after the ], keep
 330                                  * it.  We don't bother with finding the end of
 331                                  * the space, since we later normalize it
 332                                  * anyway.
 333                                  */
 334                                 if (isspace(subject->buf[at]))
 335                                         at += 1;
 336                         }
 337                         continue;
 338                 }
 339                 break;
 340         }
 341         strbuf_trim(subject);
 342 }
 343
 344 #define MAX_HDR_PARSED 10
 345 static const char *header[MAX_HDR_PARSED] = {
 346         "From","Subject","Date",
 347 };
 348
 349 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 350 {
 351         int len = strlen(hdr);
 352         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 353                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 354 }
 355
 356 static int is_format_patch_separator(const char *line, int len)
 357 {
 358         static const char SAMPLE[] =
 359                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 360         const char *cp;
 361
 362         if (len != strlen(SAMPLE))
 363                 return 0;
 364         if (!skip_prefix(line, "From ", &cp))
 365                 return 0;
 366         if (strspn(cp, "0123456789abcdef") != 40)
 367                 return 0;
 368         cp += 40;
 369         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 370 }
 371
 372 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 373 {
 374         const char *in = q_seg->buf;
 375         int c;
 376         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 377         strbuf_init(out, q_seg->len);
 378
 379         while ((c = *in++) != 0) {
 380                 if (c == '=') {
 381                         int ch, d = *in;
 382                         if (d == '\n' || !d)
 383                                 break; /* drop trailing newline */
 384                         ch = hex2chr(in);
 385                         if (ch >= 0) {
 386                                 strbuf_addch(out, ch);
 387                                 in += 2;
 388                                 continue;
 389                         }
 390                         /* garbage -- fall through */
 391                 }
 392                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 393                         c = 0x20;
 394                 strbuf_addch(out, c);
 395         }
 396         return out;
 397 }
 398
 399 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 400 {
 401         /* Decode in..ep, possibly in-place to ot */
 402         int c, pos = 0, acc = 0;
 403         const char *in = b_seg->buf;
 404         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 405         strbuf_init(out, b_seg->len);
 406
 407         while ((c = *in++) != 0) {
 408                 if (c == '+')
 409                         c = 62;
 410                 else if (c == '/')
 411                         c = 63;
 412                 else if ('A' <= c && c <= 'Z')
 413                         c -= 'A';
 414                 else if ('a' <= c && c <= 'z')
 415                         c -= 'a' - 26;
 416                 else if ('0' <= c && c <= '9')
 417                         c -= '0' - 52;
 418                 else
 419                         continue; /* garbage */
 420                 switch (pos++) {
 421                 case 0:
 422                         acc = (c << 2);
 423                         break;
 424                 case 1:
 425                         strbuf_addch(out, (acc | (c >> 4)));
 426                         acc = (c & 15) << 4;
 427                         break;
 428                 case 2:
 429                         strbuf_addch(out, (acc | (c >> 2)));
 430                         acc = (c & 3) << 6;
 431                         break;
 432                 case 3:
 433                         strbuf_addch(out, (acc | c));
 434                         acc = pos = 0;
 435                         break;
 436                 }
 437         }
 438         return out;
 439 }
 440
 441 static int convert_to_utf8(struct mailinfo *mi,
 442                            struct strbuf *line, const char *charset)
 443 {
 444         char *out;
 445
 446         if (!mi->metainfo_charset || !charset || !*charset)
 447                 return 0;
 448
 449         if (same_encoding(mi->metainfo_charset, charset))
 450                 return 0;
 451         out = reencode_string(line->buf, mi->metainfo_charset, charset);
 452         if (!out) {
 453                 mi->input_error = -1;
 454                 return error("cannot convert from %s to %s",
 455                              charset, mi->metainfo_charset);
 456         }
 457         strbuf_attach(line, out, strlen(out), strlen(out));
 458         return 0;
 459 }
 460
 461 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 462 {
 463         char *in, *ep, *cp;
 464         struct strbuf outbuf = STRBUF_INIT, *dec;
 465         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 466         int found_error = 1; /* pessimism */
 467
 468         in = it->buf;
 469         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 470                 int encoding;
 471                 strbuf_reset(&charset_q);
 472                 strbuf_reset(&piecebuf);
 473
 474                 if (in != ep) {
 475                         /*
 476                          * We are about to process an encoded-word
 477                          * that begins at ep, but there is something
 478                          * before the encoded word.
 479                          */
 480                         char *scan;
 481                         for (scan = in; scan < ep; scan++)
 482                                 if (!isspace(*scan))
 483                                         break;
 484
 485                         if (scan != ep || in == it->buf) {
 486                                 /*
 487                                  * We should not lose that "something",
 488                                  * unless we have just processed an
 489                                  * encoded-word, and there is only LWS
 490                                  * before the one we are about to process.
 491                                  */
 492                                 strbuf_add(&outbuf, in, ep - in);
 493                         }
 494                 }
 495                 /* E.g.
 496                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 497                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 498                  */
 499                 ep += 2;
 500
 501                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 502                         goto release_return;
 503
 504                 if (cp + 3 - it->buf > it->len)
 505                         goto release_return;
 506                 strbuf_add(&charset_q, ep, cp - ep);
 507
 508                 encoding = cp[1];
 509                 if (!encoding || cp[2] != '?')
 510                         goto release_return;
 511                 ep = strstr(cp + 3, "?=");
 512                 if (!ep)
 513                         goto release_return;
 514                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 515                 switch (tolower(encoding)) {
 516                 default:
 517                         goto release_return;
 518                 case 'b':
 519                         dec = decode_b_segment(&piecebuf);
 520                         break;
 521                 case 'q':
 522                         dec = decode_q_segment(&piecebuf, 1);
 523                         break;
 524                 }
 525                 if (convert_to_utf8(mi, dec, charset_q.buf))
 526                         goto release_return;
 527
 528                 strbuf_addbuf(&outbuf, dec);
 529                 strbuf_release(dec);
 530                 free(dec);
 531                 in = ep + 2;
 532         }
 533         strbuf_addstr(&outbuf, in);
 534         strbuf_reset(it);
 535         strbuf_addbuf(it, &outbuf);
 536         found_error = 0;
 537 release_return:
 538         strbuf_release(&outbuf);
 539         strbuf_release(&charset_q);
 540         strbuf_release(&piecebuf);
 541
 542         if (found_error)
 543                 mi->input_error = -1;
 544 }
 545
 546 static int check_header(struct mailinfo *mi,
 547                         const struct strbuf *line,
 548                         struct strbuf *hdr_data[], int overwrite)
 549 {
 550         int i, ret = 0, len;
 551         struct strbuf sb = STRBUF_INIT;
 552
 553         /* search for the interesting parts */
 554         for (i = 0; header[i]; i++) {
 555                 int len = strlen(header[i]);
 556                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 557                         /* Unwrap inline B and Q encoding, and optionally
 558                          * normalize the meta information to utf8.
 559                          */
 560                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 561                         decode_header(mi, &sb);
 562                         handle_header(&hdr_data[i], &sb);
 563                         ret = 1;
 564                         goto check_header_out;
 565                 }
 566         }
 567
 568         /* Content stuff */
 569         if (cmp_header(line, "Content-Type")) {
 570                 len = strlen("Content-Type: ");
 571                 strbuf_add(&sb, line->buf + len, line->len - len);
 572                 decode_header(mi, &sb);
 573                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 574                 handle_content_type(mi, &sb);
 575                 ret = 1;
 576                 goto check_header_out;
 577         }
 578         if (cmp_header(line, "Content-Transfer-Encoding")) {
 579                 len = strlen("Content-Transfer-Encoding: ");
 580                 strbuf_add(&sb, line->buf + len, line->len - len);
 581                 decode_header(mi, &sb);
 582                 handle_content_transfer_encoding(mi, &sb);
 583                 ret = 1;
 584                 goto check_header_out;
 585         }
 586         if (cmp_header(line, "Message-Id")) {
 587                 len = strlen("Message-Id: ");
 588                 strbuf_add(&sb, line->buf + len, line->len - len);
 589                 decode_header(mi, &sb);
 590                 if (mi->add_message_id)
 591                         mi->message_id = strbuf_detach(&sb, NULL);
 592                 ret = 1;
 593                 goto check_header_out;
 594         }
 595
 596 check_header_out:
 597         strbuf_release(&sb);
 598         return ret;
 599 }
 600
 601 /*
 602  * Returns 1 if the given line or any line beginning with the given line is an
 603  * in-body header (that is, check_header will succeed when passed
 604  * mi->s_hdr_data).
 605  */
 606 static int is_inbody_header(const struct mailinfo *mi,
 607                             const struct strbuf *line)
 608 {
 609         int i;
 610         for (i = 0; header[i]; i++)
 611                 if (!mi->s_hdr_data[i] && cmp_header(line, header[i]))
 612                         return 1;
 613         return 0;
 614 }
 615
 616 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 617 {
 618         struct strbuf *ret;
 619
 620         switch (mi->transfer_encoding) {
 621         case TE_QP:
 622                 ret = decode_q_segment(line, 0);
 623                 break;
 624         case TE_BASE64:
 625                 ret = decode_b_segment(line);
 626                 break;
 627         case TE_DONTCARE:
 628         default:
 629                 return;
 630         }
 631         strbuf_reset(line);
 632         strbuf_addbuf(line, ret);
 633         strbuf_release(ret);
 634         free(ret);
 635 }
 636
 637 static inline int patchbreak(const struct strbuf *line)
 638 {
 639         size_t i;
 640
 641         /* Beginning of a "diff -" header? */
 642         if (starts_with(line->buf, "diff -"))
 643                 return 1;
 644
 645         /* CVS "Index: " line? */
 646         if (starts_with(line->buf, "Index: "))
 647                 return 1;
 648
 649         /*
 650          * "--- <filename>" starts patches without headers
 651          * "---<sp>*" is a manual separator
 652          */
 653         if (line->len < 4)
 654                 return 0;
 655
 656         if (starts_with(line->buf, "---")) {
 657                 /* space followed by a filename? */
 658                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 659                         return 1;
 660                 /* Just whitespace? */
 661                 for (i = 3; i < line->len; i++) {
 662                         unsigned char c = line->buf[i];
 663                         if (c == '\n')
 664                                 return 1;
 665                         if (!isspace(c))
 666                                 break;
 667                 }
 668                 return 0;
 669         }
 670         return 0;
 671 }
 672
 673 static int is_scissors_line(const char *line)
 674 {
 675         const char *c;
 676         int scissors = 0, gap = 0;
 677         const char *first_nonblank = NULL, *last_nonblank = NULL;
 678         int visible, perforation = 0, in_perforation = 0;
 679
 680         for (c = line; *c; c++) {
 681                 if (isspace(*c)) {
 682                         if (in_perforation) {
 683                                 perforation++;
 684                                 gap++;
 685                         }
 686                         continue;
 687                 }
 688                 last_nonblank = c;
 689                 if (first_nonblank == NULL)
 690                         first_nonblank = c;
 691                 if (*c == '-') {
 692                         in_perforation = 1;
 693                         perforation++;
 694                         continue;
 695                 }
 696                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 697                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 698                         in_perforation = 1;
 699                         perforation += 2;
 700                         scissors += 2;
 701                         c++;
 702                         continue;
 703                 }
 704                 in_perforation = 0;
 705         }
 706
 707         /*
 708          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 709          * Even though there can be arbitrary cruft on the same line
 710          * (e.g. "cut here"), in order to avoid misidentification, the
 711          * perforation must occupy more than a third of the visible
 712          * width of the line, and dashes and scissors must occupy more
 713          * than half of the perforation.
 714          */
 715
 716         if (first_nonblank && last_nonblank)
 717                 visible = last_nonblank - first_nonblank + 1;
 718         else
 719                 visible = 0;
 720         return (scissors && 8 <= visible &&
 721                 visible < perforation * 3 &&
 722                 gap * 2 < perforation);
 723 }
 724
 725 static void flush_inbody_header_accum(struct mailinfo *mi)
 726 {
 727         if (!mi->inbody_header_accum.len)
 728                 return;
 729         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 730                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 731         strbuf_reset(&mi->inbody_header_accum);
 732 }
 733
 734 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 735 {
 736         if (mi->inbody_header_accum.len &&
 737             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 738                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 739                         /*
 740                          * This is a scissors line; do not consider this line
 741                          * as a header continuation line.
 742                          */
 743                         flush_inbody_header_accum(mi);
 744                         return 0;
 745                 }
 746                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 747                 strbuf_addbuf(&mi->inbody_header_accum, line);
 748                 return 1;
 749         }
 750
 751         flush_inbody_header_accum(mi);
 752
 753         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 754                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 755         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 756                 int i;
 757                 for (i = 0; header[i]; i++)
 758                         if (!strcmp("Subject", header[i])) {
 759                                 handle_header(&mi->s_hdr_data[i], line);
 760                                 return 1;
 761                         }
 762                 return 0;
 763         }
 764         if (is_inbody_header(mi, line)) {
 765                 strbuf_addbuf(&mi->inbody_header_accum, line);
 766                 return 1;
 767         }
 768         return 0;
 769 }
 770
 771 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 772 {
 773         assert(!mi->filter_stage);
 774
 775         if (mi->header_stage) {
 776                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 777                         if (mi->inbody_header_accum.len) {
 778                                 flush_inbody_header_accum(mi);
 779                                 mi->header_stage = 0;
 780                         }
 781                         return 0;
 782                 }
 783         }
 784
 785         if (mi->use_inbody_headers && mi->header_stage) {
 786                 mi->header_stage = check_inbody_header(mi, line);
 787                 if (mi->header_stage)
 788                         return 0;
 789         } else
 790                 /* Only trim the first (blank) line of the commit message
 791                  * when ignoring in-body headers.
 792                  */
 793                 mi->header_stage = 0;
 794
 795         /* normalize the log message to UTF-8. */
 796         if (convert_to_utf8(mi, line, mi->charset.buf))
 797                 return 0; /* mi->input_error already set */
 798
 799         if (mi->use_scissors && is_scissors_line(line->buf)) {
 800                 int i;
 801
 802                 strbuf_setlen(&mi->log_message, 0);
 803                 mi->header_stage = 1;
 804
 805                 /*
 806                  * We may have already read "secondary headers"; purge
 807                  * them to give ourselves a clean restart.
 808                  */
 809                 for (i = 0; header[i]; i++) {
 810                         if (mi->s_hdr_data[i])
 811                                 strbuf_release(mi->s_hdr_data[i]);
 812                         mi->s_hdr_data[i] = NULL;
 813                 }
 814                 return 0;
 815         }
 816
 817         if (patchbreak(line)) {
 818                 if (mi->message_id)
 819                         strbuf_addf(&mi->log_message,
 820                                     "Message-Id: %s\n", mi->message_id);
 821                 return 1;
 822         }
 823
 824         strbuf_addbuf(&mi->log_message, line);
 825         return 0;
 826 }
 827
 828 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 829 {
 830         fwrite(line->buf, 1, line->len, mi->patchfile);
 831         mi->patch_lines++;
 832 }
 833
 834 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 835 {
 836         switch (mi->filter_stage) {
 837         case 0:
 838                 if (!handle_commit_msg(mi, line))
 839                         break;
 840                 mi->filter_stage++;
 841                 /* fallthrough */
 842         case 1:
 843                 handle_patch(mi, line);
 844                 break;
 845         }
 846 }
 847
 848 static int is_rfc2822_header(const struct strbuf *line)
 849 {
 850         /*
 851          * The section that defines the loosest possible
 852          * field name is "3.6.8 Optional fields".
 853          *
 854          * optional-field = field-name ":" unstructured CRLF
 855          * field-name = 1*ftext
 856          * ftext = %d33-57 / %59-126
 857          */
 858         int ch;
 859         char *cp = line->buf;
 860
 861         /* Count mbox From headers as headers */
 862         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 863                 return 1;
 864
 865         while ((ch = *cp++)) {
 866                 if (ch == ':')
 867                         return 1;
 868                 if ((33 <= ch && ch <= 57) ||
 869                     (59 <= ch && ch <= 126))
 870                         continue;
 871                 break;
 872         }
 873         return 0;
 874 }
 875
 876 static int read_one_header_line(struct strbuf *line, FILE *in)
 877 {
 878         struct strbuf continuation = STRBUF_INIT;
 879
 880         /* Get the first part of the line. */
 881         if (strbuf_getline_lf(line, in))
 882                 return 0;
 883
 884         /*
 885          * Is it an empty line or not a valid rfc2822 header?
 886          * If so, stop here, and return false ("not a header")
 887          */
 888         strbuf_rtrim(line);
 889         if (!line->len || !is_rfc2822_header(line)) {
 890                 /* Re-add the newline */
 891                 strbuf_addch(line, '\n');
 892                 return 0;
 893         }
 894
 895         /*
 896          * Now we need to eat all the continuation lines..
 897          * Yuck, 2822 header "folding"
 898          */
 899         for (;;) {
 900                 int peek;
 901
 902                 peek = fgetc(in);
 903                 if (peek == EOF)
 904                         break;
 905                 ungetc(peek, in);
 906                 if (peek != ' ' && peek != '\t')
 907                         break;
 908                 if (strbuf_getline_lf(&continuation, in))
 909                         break;
 910                 continuation.buf[0] = ' ';
 911                 strbuf_rtrim(&continuation);
 912                 strbuf_addbuf(line, &continuation);
 913         }
 914         strbuf_release(&continuation);
 915
 916         return 1;
 917 }
 918
 919 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 920 {
 921         while (!strbuf_getline_lf(line, mi->input)) {
 922                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 923                         return 1;
 924         }
 925         return 0;
 926 }
 927
 928 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 929 {
 930         struct strbuf newline = STRBUF_INIT;
 931
 932         strbuf_addch(&newline, '\n');
 933 again:
 934         if (line->len >= (*(mi->content_top))->len + 2 &&
 935             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 936                 /* we hit an end boundary */
 937                 /* pop the current boundary off the stack */
 938                 strbuf_release(*(mi->content_top));
 939                 FREE_AND_NULL(*(mi->content_top));
 940
 941                 /* technically won't happen as is_multipart_boundary()
 942                    will fail first.  But just in case..
 943                  */
 944                 if (--mi->content_top < mi->content) {
 945                         error("Detected mismatched boundaries, can't recover");
 946                         mi->input_error = -1;
 947                         mi->content_top = mi->content;
 948                         strbuf_release(&newline);
 949                         return 0;
 950                 }
 951                 handle_filter(mi, &newline);
 952                 strbuf_release(&newline);
 953                 if (mi->input_error)
 954                         return 0;
 955
 956                 /* skip to the next boundary */
 957                 if (!find_boundary(mi, line))
 958                         return 0;
 959                 goto again;
 960         }
 961
 962         /* set some defaults */
 963         mi->transfer_encoding = TE_DONTCARE;
 964         strbuf_reset(&mi->charset);
 965
 966         /* slurp in this section's info */
 967         while (read_one_header_line(line, mi->input))
 968                 check_header(mi, line, mi->p_hdr_data, 0);
 969
 970         strbuf_release(&newline);
 971         /* replenish line */
 972         if (strbuf_getline_lf(line, mi->input))
 973                 return 0;
 974         strbuf_addch(line, '\n');
 975         return 1;
 976 }
 977
 978 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
 979                                  struct strbuf *prev)
 980 {
 981         size_t len = line->len;
 982         const char *rest;
 983
 984         if (!mi->format_flowed) {
 985                 handle_filter(mi, line);
 986                 return;
 987         }
 988
 989         if (line->buf[len - 1] == '\n') {
 990                 len--;
 991                 if (len && line->buf[len - 1] == '\r')
 992                         len--;
 993         }
 994
 995         /* Keep signature separator as-is. */
 996         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
 997                 if (prev->len) {
 998                         handle_filter(mi, prev);
 999                         strbuf_reset(prev);
1000                 }
1001                 handle_filter(mi, line);
1002                 return;
1003         }
1004
1005         /* Unstuff space-stuffed line. */
1006         if (len && line->buf[0] == ' ') {
1007                 strbuf_remove(line, 0, 1);
1008                 len--;
1009         }
1010
1011         /* Save flowed line for later, but without the soft line break. */
1012         if (len && line->buf[len - 1] == ' ') {
1013                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1014                 return;
1015         }
1016
1017         /* Prepend any previous partial lines */
1018         strbuf_insert(line, 0, prev->buf, prev->len);
1019         strbuf_reset(prev);
1020
1021         handle_filter(mi, line);
1022 }
1023
1024 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1025 {
1026         struct strbuf prev = STRBUF_INIT;
1027
1028         /* Skip up to the first boundary */
1029         if (*(mi->content_top)) {
1030                 if (!find_boundary(mi, line))
1031                         goto handle_body_out;
1032         }
1033
1034         do {
1035                 /* process any boundary lines */
1036                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1037                         /* flush any leftover */
1038                         if (prev.len) {
1039                                 handle_filter(mi, &prev);
1040                                 strbuf_reset(&prev);
1041                         }
1042                         if (!handle_boundary(mi, line))
1043                                 goto handle_body_out;
1044                 }
1045
1046                 /* Unwrap transfer encoding */
1047                 decode_transfer_encoding(mi, line);
1048
1049                 switch (mi->transfer_encoding) {
1050                 case TE_BASE64:
1051                 case TE_QP:
1052                 {
1053                         struct strbuf **lines, **it, *sb;
1054
1055                         /* Prepend any previous partial lines */
1056                         strbuf_insert(line, 0, prev.buf, prev.len);
1057                         strbuf_reset(&prev);
1058
1059                         /*
1060                          * This is a decoded line that may contain
1061                          * multiple new lines.  Pass only one chunk
1062                          * at a time to handle_filter()
1063                          */
1064                         lines = strbuf_split(line, '\n');
1065                         for (it = lines; (sb = *it); it++) {
1066                                 if (*(it + 1) == NULL) /* The last line */
1067                                         if (sb->buf[sb->len - 1] != '\n') {
1068                                                 /* Partial line, save it for later. */
1069                                                 strbuf_addbuf(&prev, sb);
1070                                                 break;
1071                                         }
1072                                 handle_filter_flowed(mi, sb, &prev);
1073                         }
1074                         /*
1075                          * The partial chunk is saved in "prev" and will be
1076                          * appended by the next iteration of read_line_with_nul().
1077                          */
1078                         strbuf_list_free(lines);
1079                         break;
1080                 }
1081                 default:
1082                         handle_filter_flowed(mi, line, &prev);
1083                 }
1084
1085                 if (mi->input_error)
1086                         break;
1087         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1088
1089         if (prev.len)
1090                 handle_filter(mi, &prev);
1091
1092         flush_inbody_header_accum(mi);
1093
1094 handle_body_out:
1095         strbuf_release(&prev);
1096 }
1097
1098 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1099 {
1100         const char *sp = data->buf;
1101         while (1) {
1102                 char *ep = strchr(sp, '\n');
1103                 int len;
1104                 if (!ep)
1105                         len = strlen(sp);
1106                 else
1107                         len = ep - sp;
1108                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1109                 if (!ep)
1110                         break;
1111                 sp = ep + 1;
1112         }
1113 }
1114
1115 static void handle_info(struct mailinfo *mi)
1116 {
1117         struct strbuf *hdr;
1118         int i;
1119
1120         for (i = 0; header[i]; i++) {
1121                 /* only print inbody headers if we output a patch file */
1122                 if (mi->patch_lines && mi->s_hdr_data[i])
1123                         hdr = mi->s_hdr_data[i];
1124                 else if (mi->p_hdr_data[i])
1125                         hdr = mi->p_hdr_data[i];
1126                 else
1127                         continue;
1128
1129                 if (!strcmp(header[i], "Subject")) {
1130                         if (!mi->keep_subject) {
1131                                 cleanup_subject(mi, hdr);
1132                                 cleanup_space(hdr);
1133                         }
1134                         output_header_lines(mi->output, "Subject", hdr);
1135                 } else if (!strcmp(header[i], "From")) {
1136                         cleanup_space(hdr);
1137                         handle_from(mi, hdr);
1138                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1139                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1140                 } else {
1141                         cleanup_space(hdr);
1142                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1143                 }
1144         }
1145         fprintf(mi->output, "\n");
1146 }
1147
1148 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1149 {
1150         FILE *cmitmsg;
1151         int peek;
1152         struct strbuf line = STRBUF_INIT;
1153
1154         cmitmsg = fopen(msg, "w");
1155         if (!cmitmsg) {
1156                 perror(msg);
1157                 return -1;
1158         }
1159         mi->patchfile = fopen(patch, "w");
1160         if (!mi->patchfile) {
1161                 perror(patch);
1162                 fclose(cmitmsg);
1163                 return -1;
1164         }
1165
1166         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1167         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1168
1169         do {
1170                 peek = fgetc(mi->input);
1171                 if (peek == EOF) {
1172                         fclose(cmitmsg);
1173                         return error("empty patch: '%s'", patch);
1174                 }
1175         } while (isspace(peek));
1176         ungetc(peek, mi->input);
1177
1178         /* process the email header */
1179         while (read_one_header_line(&line, mi->input))
1180                 check_header(mi, &line, mi->p_hdr_data, 1);
1181
1182         handle_body(mi, &line);
1183         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1184         fclose(cmitmsg);
1185         fclose(mi->patchfile);
1186
1187         handle_info(mi);
1188         strbuf_release(&line);
1189         return mi->input_error;
1190 }
1191
1192 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1193 {
1194         struct mailinfo *mi = mi_;
1195
1196         if (!starts_with(var, "mailinfo."))
1197                 return git_default_config(var, value, NULL);
1198         if (!strcmp(var, "mailinfo.scissors")) {
1199                 mi->use_scissors = git_config_bool(var, value);
1200                 return 0;
1201         }
1202         /* perhaps others here */
1203         return 0;
1204 }
1205
1206 void setup_mailinfo(struct mailinfo *mi)
1207 {
1208         memset(mi, 0, sizeof(*mi));
1209         strbuf_init(&mi->name, 0);
1210         strbuf_init(&mi->email, 0);
1211         strbuf_init(&mi->charset, 0);
1212         strbuf_init(&mi->log_message, 0);
1213         strbuf_init(&mi->inbody_header_accum, 0);
1214         mi->header_stage = 1;
1215         mi->use_inbody_headers = 1;
1216         mi->content_top = mi->content;
1217         git_config(git_mailinfo_config, mi);
1218 }
1219
1220 void clear_mailinfo(struct mailinfo *mi)
1221 {
1222         int i;
1223
1224         strbuf_release(&mi->name);
1225         strbuf_release(&mi->email);
1226         strbuf_release(&mi->charset);
1227         strbuf_release(&mi->inbody_header_accum);
1228         free(mi->message_id);
1229
1230         if (mi->p_hdr_data)
1231                 for (i = 0; mi->p_hdr_data[i]; i++)
1232                         strbuf_release(mi->p_hdr_data[i]);
1233         free(mi->p_hdr_data);
1234         if (mi->s_hdr_data)
1235                 for (i = 0; mi->s_hdr_data[i]; i++)
1236                         strbuf_release(mi->s_hdr_data[i]);
1237         free(mi->s_hdr_data);
1238
1239         while (mi->content < mi->content_top) {
1240                 free(*(mi->content_top));
1241                 mi->content_top--;
1242         }
1243
1244         strbuf_release(&mi->log_message);
1245 }